Question on mod_sysfs_init and kobject_put in error handling code.

[Ben Greear <greearb@candelatech.com>]

I'm seeing a crash (on hacked 3.9.3+ kernels).  It's rare, but in a kernel
larded down with debugging, we are having some luck reproducing it.

Please note, this kernel is running a fair amount of my patches, so it could
be my bug.  We did not see this before 3.9.3, as far as we know..but I have not
tried bisecting this yet.

The crash happens on startup, and I see this splat or similar:


microcode: CPU3 sig=0x20652, pf=0x10, revision=0x9
microcode: CPU3 updated to revision 0xd, date = 2011-09-01
microcode: Microcode Update Driver: v2.00 <tigran@aivazian.fsnet.co.uk>, Peter Oruba
e1000e 0000:01:00.0 eth0: MAC: 3, PHY: 8, PBA No: FFFFFF-0FF
kvm: disabled by bios
ieee80211 phy0: Atheros AR9300 Rev:3 mem=0xffffc90023a40000, irq=18
kvm_intel: module is already loaded
BUG: unable to handle kernel paging request at ffffffffa08e8700
IP: [<ffffffff813018e3>] kset_find_obj+0x23/0x7a
PGD 1a0f067 PUD 1a10063 PMD 21e9ae067 PTE 0
Oops: 0000 [#1] PREEMPT SMP
Modules linked in: acpi_cpufreq(+) kvm_intel(+) mperf intel_powerclamp kvm cdc_acm microcode serio_raw pcspkr snd_hda_codec_realtek ath9k(+) ath9k_common 
ath9k_hw ath e1000e ptp snd_hda_intel mac80211 snd_hda_codec snd_hwdep snd_seq i2c_i801 snd_seq_device lpc_ich pps_core cfg80211 snd_pcm snd_page_alloc 
snd_timer snd soundcore parport_pc parport uinput ipv6 i915 video i2c_algo_bit drm_kms_helper drm i2c_core
CPU 0
Pid: 498, comm: udevd Not tainted 3.9.4+ #3 To be filled by O.E.M. To be filled by O.E.M./To be filled by O.E.M.
RIP: 0010:[<ffffffff813018e3>]  [<ffffffff813018e3>] kset_find_obj+0x23/0x7a
RSP: 0018:ffff88021f145d68  EFLAGS: 00010293
RAX: ffffffffa08e8708 RBX: ffffffffa08e8700 RCX: 000000000000908f
RDX: 0000000000008f61 RSI: ffffffffa0958029 RDI: ffff88021527a691
RBP: ffff88021f145d88 R08: 0000000000000000 R09: ffff88021f145c78
R10: 0000000000000000 R11: 0000000000000000 R12: ffff880221cd8410
R13: ffff880221cd8420 R14: ffffffffa0958028 R15: ffffffffa0958028
FS:  00007fa594c7f840(0000) GS:ffff88022bc00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: ffffffffa08e8700 CR3: 00000002150a0000 CR4: 00000000000007f0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Process udevd (pid: 498, threadinfo ffff88021f144000, task ffff880215550000)
Stack:
  ffffffffa0958010 ffff88021f145ef8 0000000000000000 ffffffffa0958028
  ffff88021f145df8 ffffffff810f2e06 ffffffff81a51430 0000000000000246
  0000000180007fff ffffffffa09575e8 ffff880200000001 ffffffffa09575e8
Call Trace:
  [<ffffffff810f2e06>] mod_sysfs_setup+0x52/0x522
  [<ffffffff810f46be>] load_module+0x13e8/0x15cc
  [<ffffffff8131dd69>] ? ddebug_dyndbg_boot_param_cb+0x45/0x45
  [<ffffffff810f4a52>] sys_init_module+0xfd/0x103
  [<ffffffff815f13d9>] system_call_fastpath+0x16/0x1b
Code: ff 9d e8 ff 5e 5b c9 c3 55 48 89 e5 41 56 49 89 f6 41 55 4c 8d 6f 10 41 54 49 89 fc 4c 89 ef 53 e8 af 88 2e 00 49 8b 1c 24 eb 34 <48> 8b 3b 48 85 ff 74 28 
4c 89 f6 e8 1f 55 00 00 85 c0 75 1c 8b
RIP  [<ffffffff813018e3>] kset_find_obj+0x23/0x7a
  RSP <ffff88021f145d68>
CR2: ffffffffa08e8700
---[ end trace cc75890eca7ff0aa ]---


While poking around the code, I wonder if the kobject put is correct below?

In both cases I've reproduced I see the error about 'module is already loaded'
printed in the logs right before the crash.  It was a different module name each time.

The kset_find_obj plays some tricks where it only grabs a reference
sometimes...maybe the kobject_put needs some similar conditions?


static int mod_sysfs_init(struct module *mod)
{
	int err;
	struct kobject *kobj;

	if (!module_sysfs_initialized) {
		printk(KERN_ERR "%s: module sysfs not initialized\n",
		       mod->name);
		err = -EINVAL;
		goto out;
	}

	kobj = kset_find_obj(module_kset, mod->name);
	if (kobj) {
		printk(KERN_ERR "%s: module is already loaded\n", mod->name);
		kobject_put(kobj);
		err = -EINVAL;
		goto out;
	}

-- 
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc  http://www.candelatech.com